原文：
This post found its way because I couldnt find the answer to one of the questions I asked on StackOverflow (http://stackoverflow.com/questions/18546066/pixel-perfect-collision-detection-in-cocos2dx) and thought there would be others like me in search for an answer.
Collision detection is an integral part of almost all games. It is used to find when a bullet hits an enemy or when you bump into a wall etc.
There are many different requirements when we do collision detection and depending on our game we choose one of the many detection techniques.
The default Collision detection mechanism used by games and provided in almost all game engines and frameworks is a “Bounding Box” collision.
Simply put, a “Bounding Box” collision detection system the sprites/objects being checked for collision are treated as the smallest rectangle which completely engulfs them. Then these two Boxes are checked if they are colliding with each other.
But sometimes this very simple collision detection system is not accurate. Specially when we use sprites with alpha values (mostly png files) or when our objects are rotated by some angles. See the image below:
Pixel – Perfect collision detection is a system where we check if the objects concerned are actually colliding rather than just being part of a bounding box which is bigger than their size. WARNING: This system though more accurate is obviously more performance intensive and hence depending on your game requirements choose wisely about which of the different systems you want to use.
TIP: This system though written specially for Cocos2d-x framework can be easily understood and implemented for any language/framework you are using.
So its time to get our hands dirty,
We are going to develop a Singleton Class for collision detection and just plug and play this in any project we are doing.
Things used:
1. Singleton Class – CollisionDetection
2. Opengl Vertex and Fragment Shaders
3. CCRenderTexture Class – Cocos2d-x
Theory:
1. Create a CCRenderTexture which is going to serve as a secondary draw buffer.
2. We first do a simple collision detection (Bounding Box) to check if the two sprite’s bounds are colliding
3. If step 2 is a success then we are going to draw the two concerned objects in our secondary buffer we created in step 1. (We are going to set its visibility to false, so that even though we draw something, nothing will we visible to the end user)
4. Using openGL fragment shaders we are going to draw one of the objects completely RED and the other completely BLUE! 
5. Using another of openGL functionality glReadPixels we are going to read the pixels data of all the pixels in the Rectangular area (Intersection area) of the bounding box collision
6. We are then going to loop through all the pixel values and check if a single pixel has BOTH the RED and the BLUE pixels. If they have then the objects are actually colliding or else not.
Now here is the code for the above steps. I have commented the code for you to understand what is going on. If there are any questions please leave in the comments and I will try and answer to the best of my knowledge
CollisionDetection.h
CollisionDetection.cpp
SolidColorShader.fsh
SolidVertexShader.vsh
For using the Collision Detection Class:
1. Initialize the CCRenderTexture object
_rt = CCRenderTexture::create(visibleSize.width * 2, visibleSize.height * 2); _rt->setPosition(ccp(visibleSize.width, visibleSize.height)); _rt->retain(); _rt->setVisible(false);
2. Call the Singleton function whenever collision detection required
if (CollisionDetection::GetInstance()->areTheSpritesColliding(pSprite, pCurrentSpriteToDrag, true, _rt)) {
//Code here
}
后记：
为了方便使用，我略去了代码展示部分，把代码统一放到压缩包里并提供下载。
问题：
原文中有一种写法在vs2012中会报错，解决方案如下：
原写法：
spr1->setBlendFunc((ccBlendFunc){ GL_SRC_ALPHA, GL_ONE });
修改后的例子：
ccBlendFunc blend_spr1 = {GL_SRC_ALPHA, GL_ONE};
spr1->setBlendFunc(blend_spr1);